As U.S. manufacturers accelerate the digitalization of their operations, data management and standardization have become increasingly prominent industry topics.

A recent survey released by the National Association of Manufacturers (NAM) shows that more than 60% of manufacturing companies have established data management strategies or guidelines, but only 15% are able to fully follow the plan. 44% of manufacturing leaders say the amount of data they collect has doubled in just the past two years, and they expect that figure to grow another threefold by the end of this decade.

The survey also found that 86% of respondents believe effectively leveraging manufacturing data is critical, but only a quarter have high confidence that they are collecting the right data.

Prateek Kathpal, president of SymphonyAI's industrial division, noted that AI models and agentic systems rely on accurate, real-time data for every decision, so when data goes wrong at the source, it can have a cascading effect on model outputs and cause millions of dollars in losses. The company's IRIS Foundry platform is specifically designed to integrate IT, OT, and engineering data to provide AI-ready analytics.

"Machines in factories generate massive amounts of data, but that data is often trapped in siloed systems, limiting visibility and slowing down everything from root cause analysis to predictive maintenance," Kathpal said.

The gap between AI ambition and AI readiness

Although most companies are excited about AI's potential, very few are truly ready to implement it.

Jasmeet Singh, executive vice president and global head of manufacturing at Infosys, said there is a clear gap between AI ambition and AI readiness. "Just as physical infrastructure determines factory operating efficiency, data infrastructure now determines whether a company can achieve intelligent operations."

Data management platform Specright, in partnership with UserEvidence, surveyed 45 manufacturers across different industries and found that 60% still use spreadsheets like Excel or Google Sheets to manage specification data, nearly half rely on shared drives such as SharePoint or Google Drive, and 38% pull data from ERP systems not designed for specification-level complexity. Nearly 70% of respondents admitted difficulty maintaining up-to-date data, with teams losing an average of 468 hours per year on manual specification tasks.

Specright CEO Mike Boese said many teams have been working around flawed systems for years. "Changing these habits—and gaining buy-in across R&D, packaging, quality, operations, and IT—requires both a cultural shift and executive-level push, and not every company has mobilized those resources."

"If that's the foundation you're building AI on, you're not going to get intelligent results," he added. "What you'll get is 'garbage in, garbage out'—just faster."

The problem may stem from a pursuit of immediate results. Brian Zakrajsek, smart manufacturing and operations expert leader at Deloitte, said: "Boards want AI returns on data that doesn't exist yet." He added that pressure to demonstrate AI value is driving companies to fund use cases while underinvesting in foundational infrastructure.

Build vs. buy debate

Even when companies decide to tackle data management and standardization, a major obstacle remains: whether to build technology in-house or adopt third-party solutions.

Maor Farid, founder and CEO of Leo AI, a company building AI for mechanical engineering, said: "Large manufacturers almost always start with internal attempts. They have data science teams, budgets, and an understandable instinct—engineering knowledge is too sensitive and specialized to hand to outside vendors, especially with defense and medical device customers, who are the most demanding."

Singh believes manufacturers understand their products, processes, plants, and customers better than anyone, and that domain expertise is critical. "However, scaling AI and data modernization across large manufacturing enterprises also requires expertise in cloud, platform engineering, cybersecurity, data governance, AI model operations, systems integration, and change management."

Kathpal said this is driving more manufacturers toward external solutions, especially because building systems from scratch can lead to lengthy and costly deployments. Increased competition is also a key factor, because "your competitors may be deploying projects at scale and already seeing results."

The most effective use cases

Some of these results are tackling the biggest challenges that have plagued the industry for decades.

Farid pointed to one of them: "preserving and passing on knowledge—knowledge that resides in the minds of veteran employees nearing retirement, and in design decisions buried in old projects that were never documented." When this knowledge is captured in an accessible, searchable form, new employee onboarding becomes more efficient, senior staff can focus on important tasks rather than training newcomers, and companies don't lose critical data when people leave.

"When engineers can ask questions in natural language and immediately get answers from their own company data and relevant industry standards, the time savings are staggering," he added. "For regulated industries, this also brings compliance advantages, because traceable, sourced knowledge is exactly what FDA or ISO audits expect to see."

Boese said: "Sustainability reporting requirements are expanding rapidly—extended producer responsibility (EPR) enforcement alone can expose companies to fines of tens of thousands of dollars per day, per state." EPR is a policy that holds companies responsible for the end-of-life recycling and disposal of their products and packaging.

"As the regulatory environment tightens, companies with clean, connected specification data will gain a significant competitive advantage," Boese added.

Another use case is supply chain optimization. Singh said: "By better understanding suppliers, inventory, logistics, and demand, manufacturers can use AI to improve forecasting, optimize processes, and respond faster to disruptions."

Accessible data also enables more efficient predictive maintenance. Kathpal said manufacturers can get real-time alerts on equipment anomalies and reduce unplanned downtime. This approach helped one of his clients avoid $2.5 million in annual losses due to operational inefficiencies, and for very large enterprises, the figure is much higher.

"The manufacturers that win the next decade won't be those with the flashiest AI," Farid said. "They'll be those that treat data as a foundation, not an afterthought."